AWS releases
Amazon Web Services releases and Terraform AWS provider. New features, breaking changes, security advisories and deprecations - each summarised in plain English and updated continuously.
Tracking 284 AWS releases · Updated
- AWS What's New aiawsgaengineermedia ·
Amazon EC2 G7 instances with NVIDIA RTX GPUs now generally available
AWS announced the general availability of Amazon EC2 G7 instances, featuring NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. These instances offer significant performance improvements for AI inference and graphics workloads compared to G6, supporting applications like language translation, real-time graphics rendering, and game streaming. G7 instances are now available in two AWS Regions and can be purchased through various pricing models.
feature announcement - AWS What's New aiawsengineer ·
SageMaker AI adds observability for inference endpoints
Amazon SageMaker AI now offers an observability capability for generative AI inference workloads, providing real-time visibility into token performance and infrastructure health. This feature automates the collection and display of key metrics, reducing manual troubleshooting time from hours to minutes. It is designed for engineers and architects managing production AI inference, with a new CloudWatch dashboard and Grafana integration available across multiple AWS regions.
feature announcement - AWS What's New aiawsgaengineer ·
Mistral-3-14B-Instruct model available on Amazon SageMaker JumpStart
Amazon SageMaker JumpStart now offers the Mistral-3-14B-Instruct-2512 foundation model, enabling multimodal reasoning and agentic AI capabilities. This compact 14B-parameter model supports advanced AI assistants, vision-enabled applications, and native function calling, with easy deployment via SageMaker Studio or SDK. It is optimized for edge deployment and handles multilingual text and image analysis.
feature announcement - AWS What's New aiaws ·
Amazon SageMaker JumpStart adds all-MiniLM-L12-v2 for semantic search
Amazon SageMaker JumpStart now offers the all-MiniLM-L12-v2 model for semantic search and sentence similarity. This allows AWS customers to efficiently build applications like information retrieval and document clustering using dense vector representations. The model is readily deployable via SageMaker Studio or the Python SDK, ideal for production workloads needing fast, high-quality text embeddings.
feature announcement - AWS What's New aiawsgaengineer ·
Amazon Bedrock AgentCore Harness GA
AWS has launched the managed agent harness in Amazon Bedrock AgentCore, enabling faster development of functional agents. This feature provides the orchestration layer, tool execution, state management, and error recovery, previously requiring significant manual development. It allows customers to define agents via configuration rather than coding the loop, offering flexibility in model selection and integration with tools and knowledge sources. The harness is available today in all commercial AWS Regions where AgentCore is offered.
feature announcement - AWS What's New aisecurityawsengineer ·
AWS Secrets Manager adds secret safety skill to Agent Toolkit
AWS Secrets Manager now includes a secret safety skill within the Agent Toolkit for AWS's aws-core plugin, enhancing security for AI coding agents. This feature prevents sensitive secrets from being exposed to AI models or session logs during agentic workflows. It is available today for all supported agent harnesses and AWS Regions, offering a crucial security upgrade for developers building on AWS.
feature - AWS What's New aiawsengineer ·
Oracle Database@AWS adds Oracle Autonomous AI Database Serverless
Oracle Database@AWS now supports Oracle Autonomous AI Database Serverless (ADB-S), a managed Exadata service. This integration allows provisioning directly from AWS, simplifying operations for AI Transaction Processing, AI Lakehouse, AI JSON Database, and APEX workloads. ADB-S includes automated management, independent scaling, high availability, and integrates with AWS security and monitoring services, launching in US East (N. Virginia) and US West (Oregon) regions.
feature - AWS What's New aiawsdatabrickssnowflakeengineer ·
Amazon Quick Adds Autonomous Agents and Multi-Dataset Analytics
Amazon Quick has launched new features including autonomous agents, multi-dataset analytics, and a redesigned activity feed. These updates allow Quick to automate recurring tasks and provide unified analytics across multiple data sources, simplifying complex workflows for business users. The changes are particularly beneficial for those managing deals, summarizing regulatory information, or processing orders, and enable natural language querying across diverse data stores like Snowflake and relational databases. Autonomous agents can be configured for varying levels of autonomy, from approvals to goal-based execution, while the activity feed offers a conversational interface for managing updates and requests.
feature - AWS What's New aisecurityawspreviewengineer ·
AWS Security Agent adds AI-powered threat modeling
AWS Security Agent now offers AI-powered threat modeling to automatically generate threat models for applications. This feature analyzes design documents or source code, identifies threats using the STRIDE framework, and suggests mitigations, reducing manual effort and specialized expertise requirements. Available in public preview for all AWS Security Agent regions, it integrates with IDEs and can be used by developers and security teams.
feature announcement - AWS What's New aisecurityawsengineer ·
AWS Security Agent adds Kiro, Claude Code, and simulated validation
AWS Security Agent, now part of AWS Continuum, has added support for Kiro and Claude Code, allowing developers to run security scans from their IDE. The agent now simulates exploits in a sandbox to validate findings, reducing false positives and prioritizing remediation. New integrations with GitLab, GitHub Enterprise, Bitbucket, and Confluence are also included, enhancing its utility for development teams.
feature - AWS What's New aiawsgapreviewengineer ·
Amazon Bedrock AgentCore adds production agent optimization
AWS has introduced new optimization capabilities for Amazon Bedrock AgentCore, allowing agents to learn from production usage to improve continuously. These features address silent agent failures by surfacing insights into agent behavior and recommending specific, data-grounded fixes. The new tools, including failure, intent, and trajectory insights, are aimed at developers and architects building and managing AI agents, with some features available in preview and others generally available across multiple AWS Regions.
feature patch - AWS What's New aiawsgaarchitecthealthcare ·
Amazon Bedrock Guardrails Adds Automated Reasoning Checks in Sydney
Amazon Bedrock Guardrails now offers Automated Reasoning checks, utilizing formal verification to mathematically validate AI model outputs, a significant advancement over traditional sampling. This new capability enhances AI safety by reducing hallucinations, policy violations, and ambiguity, particularly benefiting regulated industries like finance and healthcare. The feature provides up to 99% accuracy with provable assurance, is now available in the Asia Pacific (Sydney) Region, and can be accessed via the Bedrock console or SDK.
feature announcement - AWS What's New aiawsgaengineer ·
Amazon Bedrock AgentCore harness now GA
AWS Bedrock AgentCore harness is now generally available, allowing teams to deploy production-grade agents quickly by defining them in configuration rather than coding the orchestration loop. This managed capability handles tool execution, state persistence, and session isolation, improving agent performance and reducing development time. Customers can also switch models mid-session and export their configurations to code for custom orchestration needs.
feature announcement - AWS What's New aisecurityawsengineer ·
Amazon Bedrock AgentCore adds Guardrails for AI agent safety
Amazon Bedrock AgentCore now integrates with Bedrock Guardrails, enhancing safety and security for AI agents in production by providing real-time defense against prompt injection and data exposure. This feature offers centralized enforcement outside agent code, with logging via AgentCore observability, and is available in select AWS regions. It requires no new infrastructure and uses consumption-based pricing for policy evaluations.
feature - AWS What's New aiawsgaengineer ·
Amazon Bedrock Managed Knowledge Base GA
Amazon Bedrock Managed Knowledge Base, a fully managed RAG service, is now generally available. This feature allows developers to build AI agents grounded in enterprise data without managing vector databases or data pipelines. It supports multiple data sources and advanced retrieval capabilities, integrating with Amazon Bedrock AgentCore for seamless agent creation. The service is available in several AWS regions.
feature announcement - AWS What's New aiawsengineer ·
Amazon S3 adds large-scale annotations for data discovery by AI and analytics tools
Amazon S3 now supports annotations, a new metadata capability allowing custom JSON, XML, or YAML context (up to 1GB per object) to be attached directly to S3 objects at scale. This feature enables AI agents and analytics tools to discover and understand data without requiring separate metadata systems, enhancing data discovery and usability. Annotations are available in all AWS Regions and can be accessed via S3 Metadata or queried using tools like Amazon Athena and SageMaker.
feature - AWS What's New aiawsengineer ·
Amazon Bedrock Guardrails adds API for granular agentic AI workflow control
Amazon Bedrock Guardrails now offers the InvokeGuardrailChecks API, enabling granular, per-request control over safeguards in agentic AI applications. This new resourceless API allows developers to apply individual content, prompt attack, and sensitive information filters at any point in an agent's loop, returning severity and confidence scores for custom actions. It addresses the challenge of scaling guardrails in iterative AI workflows by operating in a detect-only mode, offering flexibility for evolving requirements. The API is available today in select AWS Regions.
feature - AWS What's New aiawsengineer ·
Amazon S3 Vectors cuts query costs for large vector indexes
Amazon S3 Vectors has reduced data processing charges for queries on vector indexes with over 10 million vectors by up to 80%. This change lowers costs for large-scale AI, RAG, and semantic search workloads by making queries more affordable. The new pricing applies automatically and is effective immediately in all AWS Regions. While costs are reduced, distributing vectors across multiple indexes is still recommended for optimal performance.
patch - AWS What's New aiawseolengineer ·
AWS Transform adds generative AI model migration assessment
AWS Transform now supports model-to-model migration assessment for generative AI workloads, producing a plan to move from third-party providers to Amazon Bedrock. This AI-powered agent scans code, identifies models, gathers requirements, and maps them to Bedrock equivalents with cost comparisons and code changes. The feature helps consolidate AI workloads on AWS, offering enhanced security and tooling, and is available in all AWS Transform regions at no extra charge.
feature announcement - AWS What's New aiawsengineer ·
Amazon EC2 P6-B200 instances with NVIDIA Blackwell GPUs available in Asia Pacific (Mumbai)
Amazon EC2 P6-B200 instances, powered by NVIDIA Blackwell GPUs, are now available in the Asia Pacific (Mumbai) region. These instances offer up to 2x performance for AI training and inference compared to previous generations, featuring significant upgrades in GPU memory and networking capabilities. They are designed for large-scale AI workloads and are now available in multiple AWS regions, including Oregon, Virginia, Ohio, GovCloud, and Mumbai.
feature announcement